DeepSeek, the Chinese AI startup that rattled global markets earlier this year, has discreetly rolled out an upgraded version of its R1 reasoning model, according to a listing on the AI model repository Hugging Face.

The company, which made waves with its cost-efficient, open-source R1 model that outshone offerings from Meta and OpenAI, continues to disrupt the AI landscape without formal fanfare.

The enhanced DeepSeek R1, a reasoning model designed for complex, step-by-step logical tasks, ranks just behind OpenAI’s o4-mini and o3 models on LiveCodeBench, a key benchmarking platform.

The low-key release mirrors the debut of the original R1, which stunned markets by demonstrating superior performance developed at a fraction of the cost and time of its Western counterparts.

DeepSeek’s rise has sparked concerns about the spending strategies of U.S. tech giants, with the initial R1 launch triggering a sharp sell-off in stocks like Nvidia, a leader in AI infrastructure.

While major U.S. tech firms have largely recovered, DeepSeek’s advancements highlight China’s resilience in AI development despite U.S. restrictions on semiconductor exports.

This month, Chinese tech giants Baidu and Tencent also showcased efforts to optimize their AI models to circumvent U.S. chip curbs.

Meanwhile, Nvidia CEO Jensen Huang criticized U.S. export controls on Wednesday, underscoring tensions in the global AI race.

“The U.S. has based its policy on the assumption that China cannot make AI chips,” Huang said. “That assumption was always questionable, and now it’s clearly wrong.”

“The question is not whether China will have AI,” Huang added. “It already does.”